Second Language Acquisition and Processing of Norwegian DP Internal Agreement

(1)

Second Language Acquisition and Processing of Norwegian DP Internal

Agreement

Fufen Jin

Doctoral Thesis

Submitted for the Partial Fulfilment of the Requirements for the Degree of

Philosophiae Doctor

Department of Scandinavian Studies and Comparative Literature Faculty of Arts

Norwegian University of Science and Technology March 2007

(2)

(3)

NTNU 2007: 69

Printed in Norway by NTNU Trykk, Trondheim

(4)

(5)

TO MY PARENTS FENGSHAN JIN AND XUEXIANG LI, MY HUSBAND JINGYUE LI AND MY SON RUIYI LI

(6)

(7)

Abstract

The primary goal of this study is to determine whether adult second language learners ultimately achieve linguistic competence that is identical to that of native speakers, not only at the level of grammatical knowledge but also at the level of processing. This issue is addressed through investigating the production and perception of Norwegian Determiner Phrase (DP) internal agreement by endstate L2 speakers of Norwegian.

Participants in the experiment include three different L2 groups (L1 English, L1 Italian/Spanish, and L1 Chinese), and a native control group.

The target language, Norwegian, has DP-internal agreement between determiners, attributive adjectives, and nouns in terms of gender, number, and definiteness. Based on Julien’s (2005) DP model, D and α heads in Norwegian are assumed to carry uninterpretable gender, number and definiteness features. The source languages in question differ parametrically from the target language with respect to one, two or all of the uninterpretable features relevant for DP internal agreement. The informants are tested on two tasks: an online DP production task, and an auditory naming task. The purpose of the DP production task is to determine whether the uninterpretable features will be present in the interlanguage grammar regardless of status of the corresponding features in the L1s; the purpose of the auditory naming task is to examine whether L2 learners will be sensitive to the agreement cues on D when processing L2 Norwegian, and whether their sensitivity (or insensitivity) to the agreement cues is directly related to how well they produce the correct agreements.

Results from the on-line DP production task divide the L2 participants into two groups: some informants (referred to as G1 participants) performed in a nativelike way, while the others (referred to as G2 participants) showed a strong L1 effect—those uninterpretable features which are not instantiated in L1s are areas of persistent difficulty in these L2 speakers. Results from the auditory naming task show that endstate L2 speakers do not process Norwegian DP internal agreement in a nativelike fashion. While native speakers are sensitive to the agreement cues on D, L2 speakers are totally insensitive to the grammatical agreement violations in L2 grammar. This observation indicates that L2 processing is less automatic and involves less full syntactic computation than L1 processing. A comparison of G1 and G2 participants reveals a remarkably similar processing pattern — neither group shows sensitivity to the agreement cues on D, contra to what has been found in native speakers. This finding indicates that adult L2 speakers have not attained nativelike processing, even after they have attained nativelike grammatical knowledge, suggesting that certain processing mechanisms are subject to a critical period.

(8)

(9)

Preface

This thesis is submitted to the Norwegian University of Science and Technology (NTNU) for partial fulfilment of the requirements for the degree of Philosophiae Doctor.

The work referred to has been performed at the Department of Scandinavian Studies and Comparative Literature, NTNU, Trondheim, under the supervision of Professor Tor A. Åfarli.

(10)

(11)

Acknowledgements

No words can express my gratitude to my supervisor, Professor Tor A. Åfarli, who is virtually responsible for every step of my intellectual growth during my three-year Ph.D program. His vast knowledge of syntax and second language acquisition has inspired me a lot. A discussion with him is more often worth ten years of reading by myself. He read the manuscript of the thesis word by word. His careful and insightful comments helped me interpret syntactic theories correctly, present the thesis more clearly, and remove many grammatical errors.

With deepest gratitude, I thank Dr. Kristin E. Eide for her friendship, encouragement, and continuous support. I always admire her research competence and positive, cheerful personality. With her expertise on generative syntax, she has provided invaluable input to my studies. With her sunny smiles and warm friendship, she has helped me keep in good spirits even in cold, dark winter days. Her dedicated work in organizing the Språkmøter seminar and the Minimalist Syntax seminar has created a stimulating syntactic research environment at NTNU, which benefit linguistic researchers and students alike, including me.

I gratefully acknowledge Dr. Chris Wilder for allowing me to sit in his syntax course and for numerous discussions about the DP syntax in the early stage of my thesis.

Special thanks are due to Torbjørn Nordgård, Jørn Almberg and Brit Mæhlum for teaching me the methodology course.

I am grateful to our Språkmøter project research team for their cooperation and support. Our project manager, Professor Wim van Dommelen, deserves my special thanks for his technical support during the experiments and for his help with statistic analysis. I also wish to express my gratitude to the informants who participated in the experiments and patiently answered my questions, to Randi A. Nilsen for her help with the recording, to Guro Busterud and Marthe-Julie Lundeby for their help in interviewing the L2 informants, and to Rein Ove Sikveland for his help in running the experiment and checking the transcriptions.

Many researchers outside NTNU have also helped me a lot with their insights, advice, or course supervision. I am most grateful to Professor Roger Hawkins and Dr.

Ingrid Leung for their kindness and hospitality during my research visit to the University of Essex in spring, 2005. I also feel indebted to members of the second language acquisition discussion group and to my course supervisors, Ingrid Leung, Claudia Felser, Bob Borsley, Martin Atkinson and Andrew Radford. Thanks are also due to Marit Julien at Lund University for answering my questions regarding the DP structure of Norwegian. Dr. Boping Yuan at Cambridge University visited NTNU in 2006 and gave a series of lectures on second language acquisition, from which I gleaned much needed inspiration for my own study. I also thank him for his very encouraging comments on my thesis. Different parts of this work (in collaboration with Tor A. Åfarli and Wim van Dommelen) had been presented at Eurosla 16 and GALANA-2. I thank the audience for their questions and insightful comments.

I wish to thank my Norwegian language teachers Sissel Robins, Åsta Øvregaard for their instruction, and Sissel Furuseth for providing me the material in designing the Norwegian proficiency test. Special thanks go to Olaf Husby for very pleasant cooperation in writing the co-authored book, Innføring i kinesisk for nordmenn og innføring i norsk for kinesere.

(12)

I would like to express my gratitude to the administrative personnel in the HF faculty and departmental secretaries for creating a pleasant work environment. I thank my colleagues and friends, especially Ragnhild Ljosland, Annika Odland, Gunnhild Aamodt, and Mari Nygård, for being so kind to me. I also thank my Chinese friends in Trondheim, especially Naiquan Ye, Shanshan Jiang, Hua Liu, Yanjun Li, Jingdong Liu, Shaoli Yang, and Heng Yang, for their care and friendship.

My final word of gratitude is reserved for my beloved husband Dr. Jingyue Li and my son Ruiyi Li. A thousand thanks to them for making home a paradise.

(13)

3.1 Introduction 23 3.2 Some major claims of the MP 23 3.3 Syntactic structure of Norwegian DPs 27 3.4 Syntactic structure of Italian/Spanish DPs 35 3.5 Syntactic structure of English DPs 37 3.6 A note on Chinese DPs 38 3.7 Summary and a road ahead 39 4 Literature review ... 41

4.1 Introduction 41 4.2 Failed Functional Features Hypothesis (FFFH) 42 4.2.1 Hawkins and Chan (1997) ... 42

(14)

4.2.2 Hawkins (2004); Hawkins and Hattori (2006) ... 45

4.2.3 Franceschina (2001) ... 48

4.3 Full Transfer Full Access (FTFA) 49 4.3.1 Schwartz and Sprouse (1994, 1996) ... 49

4.3.2 White et al. (2004)... 51

4.3.3 Goad et al. (2003), Goad and White (2006) ... 54

4.4 Shallow Structure Hypothesis (SSH) 58 4.5 A summary 62 5 L2 production... 65

5.1 Introduction 65 5.2 The participants 66 5.2.1 The L2 participants... 66

5.2.2 Norwegian Proficiency Test (NPT)... 69

5.2.3 L1 participants ... 70

5.3 Experiment 1: Gender Assignment 71 5.3.1 Introduction ... 71

5.3.2 Method... 71

5.3.3 Results ... 72

5.4 Experiment two: On-line DP production task 74 5.4.1 Introduction ... 74

5.4.2 Predictions ... 74

5.4.3 Method... 77

5.5 Results 80 5.5.1 Rate of missed items... 80

5.5.2 L1 group ... 80

5.5.3 L2 groups... 80

5.6 Summary of findings and discussion 91 5.6.1 Summary of findings from experiment one and two ... 91

5.6.2 Predictions revisited ... 92

5.6.3 Accounting for L1-L2 differences in production ... 95

6 L1 & L2 Perception ... 99

6.1 Introduction 99 6.2 L1 and L2 feature processing 100 6.3 The experiment 105 6.3.1 Method... 106

6.3.2 Results ... 110

6.4 Summary of findings and discussion 117 6.4.1 A summary of L1 & L2 processing results... 117

6.4.2 The locus of priming effects in L1 processing ... 118

6.4.3 Agreement processing mechanisms... 119

6.4.4 L1/L2 processing differences ... 121

6.4.5 Is there an L1 transfer effects on L2 processing? ... 124

6.5 Conclusion 125 7 Conclusions and future directions... 127

(15)

7.1 Summary and conclusion 127

7.2 Contributions 128

7.3 Directions for future work 130

References... 133 Appendix... 143

(16)

List of Figures

Figure 4.1. The FFFH’s and the FTFA’s accounts for L2 ultimate attainment... 62

Figure 4.2. L2 processing model in accordance with the SSH... 63

Figure 5.1. Each L2 group’s accuracy on the gender assignment task... 74

Figure 5.2. Accuracy rate of the DPs produced by the each L2 informant ... 82

Figure 5.3. Accuracy on D and a for each L2 participant ... 82

Figure 5.4. Individual L2 group’ error rate (%) on each of the categories... 85

Figure 6.1. The L1 group’s mean RTs to concordant and discordant neuter and common gender-marking... 113

Figure 6.2. The L2 participants’ mean RTs to concordant and discordant neuter and common gender-marking... 115

List of Tables

Table 2.1. Inflectional categories and agreement in DPs of the L2 and the L1s ... 12

Table 2.2. Inflection paradigm of modified Norwegian DPs ... 18

Table 2.3. Illustrations of DP internal agreement in the L2 and L1s ... 21

Table 3.1. Parameterized uFs in the target and the source languages ... 40

Table 4.1. Chinese group’s accuracy in judging English RRCS (%) ... 43

Table 4.2. Chinese group’s accuracy in rejecting wh-island and complex NP violations ... 43

Table 4.3. French group’s accuracy in judging English RRCS (%)... 44

Table 4.4. French group’s accuracy in rejecting wh-island and complex NP violations 44 Table 4.5. Mean choice of answers corresponding to the scope of a matrix wh-word... 46

Table 4.6. Parameterized uFs in English and Spanish... 48

Table 4.7. Martin’s mean scores according to category ... 49

Table 4.8. Gender vs. number agreement errors... 49

Table 4.9. Agreement (Agr) in production by stem shape (% target-like) ... 56

Table 4.10. ERP effects to morphological violations in L2 and native speakers ... 58

Table 4.11. Response accuracy and effects to wh-dependencies in L2 learners and native speakers ... 60

Table 5.1. L2 informants’ background information and self-rated language proficiency ... 67

Table 5.2. L2 Informants’ foreign language configurations and age starting the language(s)... 68

Table 5.3. Patty’s plural marking in obligatory contexts ... 69

Table 5.4. Informants’ NPT scores (maximal points: 50) ... 70

Table 5.5. Individual informants’ accuracy on et items and en items ... 72

Table 5.6. Parameterized uninterpretable features in the L2 and the L1s ... 75

Table 5.7. Types of DPs and illustrations in the DP production task... 77

Table 5.8. Absolute accuracy, accuracy on D and a for each L2 participant ... 81

Table 5.9. Each L2 group’s errors on adjectives ... 83

Table 5.10. Each L2 group’s suppliance of /t/, /e/, ø in obligatory contexts... 83

Table 5.11. Each L2 group’s suppliance of den, det, de in obligatory contexts... 84

Table 5.12. Each L2 group’s errors on definite D ... 84

(17)

Table 5.13. Distribution of errors in each L2 group (%)... 86

Table 5.14. Rate of gender inconsistency at the individual and group level ... 87

Table 5.15. Individual G1 participants’ gender alternation pattern... 89

Table 5.16. Individual G2 participants’ gender alternation pattern... 90

Table 5.17. Phonetic forms of inflections on D, ADJ, and N in the Norwegian double definiteness construction ... 94

Table 6.1. Sample stimuli in the auditory naming task (mismatched features in bold form) ... 107

Table 6.2. Mean word frequency of the target nouns in each cell... 108

Table 6.3. Mean syllable length of the target nouns in each cell ... 108

Table 6.4. Mean word durations in the recording (in milliseconds)... 109

Table 6.5. Control group’s Mean RTs in each cell (RT in ms) ... 111

Table 6.6. The sub-control group’s mean RTs in each cell (RT in ms) ... 113

Table 6.7. The L2 informants’ mean RTs in each cell (RT in ms)... 114

Table 6.8. G1 L2 participants’ mean RTs in each cell (RT in ms)... 116

Table 6.9. G2 L2 participants’ mean RTs in each cell (RT in ms)... 116

(18)

(19)

Abbreviation list

Acc.: accusative Case ADJ: adjective

ATB: across-the-board C: complementizer

C-I: conceptual-intentional (interface) CL: classifier

COM: common gender D: prenominal determiner DEF: definiteness

DM: Distributed Morphology DP: determiner phrase

EEG: electroencephalogram ERP: event-related brain potential Exp: expression

F: feminine gender

FTFA: full transfer full access hypothesis FFFH: failed functional features hypothesis FL: the faculty of language

GB: Government and Binding theory GEN: gender

iF: interpretable feature IH: interpretability hypothesis INDEF: indefiniteness

SLA: second language acquisition LI: lexical item

L1: first language L2: second language MASC: masculine gender MP: the Minimalist Program ms: millisecond

MSIH: missing surface inflection hypothesis N: noun

NEUT: neuter gender NOM: nominative Case NS: native speaker NNS: nonnative speaker NUM: number

OEHS: onset of an empty-headed syllable PERF: perfective aspect

PL: plural

P&P: Principles and Parameters theory PPh: phonological phrase

PTH: prosodic transfer hypothesis PWd: prosodic word

Q: question

(20)

RDH: representational deficit hypothesis RRC: restrictive relative clause

RT: reaction time SG: singular

SMT: strong minimalist thesis SM: sensory-motor (interface) SSH: shallow structure hypothesis t: trace

T: tense Top: topic

uF: uninterpretable feature UG: universal grammar w: weak (inflection)

(21)

1 Overview

1.1 Introduction

A fundamental concern of second language acquisition (SLA) research is whether or not adult second language learners can ultimately reach nativelike attainment. Available SLA literature has shown that only in rare cases do adult interlanguage grammars fully converge with native grammars; by contrast, it has often been attested that interlanguage grammars diverge from native grammars in some important aspects, despite comprehensive exposure to the target language. Accounting for divergent outcomes of native speakers (NS) and nonnative speakers (NNS) thus has been taken as the central goal for the generative approach to SLA, the research paradigm within which the current study is conducted (see Hawkins 2001a, for the significance of explaining NS- NNS differences). It also constitutes the main goal of the present study.

Early attempts to account for NS-NNS differences have been centered on testing whether there is a critical period in language acquisition. The so-called Critical Period Hypothesis (CPH) was first formulated by Lenneberg (1967). The central idea of the CPH is that there is a critical period (the time span of which was initially labeled between age two and puberty) in (first and second) language acquisition; after this critical period, human capacity for learning languages declines with maturation, and nativelike linguistic competence can no longer be achieved. For researchers working within the Principles and Parameters (P&P) framework, a challenge has been to find out whether a critical period affects all linguistic domains, or none, or only certain subparts of Universal Grammar (UG). So far evidence from existing SLA studies unanimously points to no age effects on UG principles. Numerous researchers have demonstrated, typically through ‘poverty-of-the-stimulus’ phenomena,¹ that post-critical period second language (L2) learners still have access to UG principles (e.g. Kanno 1997, 1998; Pérez- Leroux and Glass 1999; Lozano 2003; Hawkins, et al. 2006, among others). For instance, in a recent study, Hawkins et al. (2006) reported that L2 learners showed

1 The poverty-of-the-stimulus argument first appeared in Chomsky (1980:34), which claims that the primary linguistic data children are exposed to are impoverished with respect to their ultimate syntactic knowledge. Based on this argument, Chomsky posits that some aspects of linguistic knowledge are innate. The poverty-of-the-stimulus phenomena have often been taken as “the hallmark for the testimony for the notion that language is inherently innate in humans” (Ritter 2001:1). See Thomas (2002) for a historical account about the emergence of the poverty-of-the-stimulus argument.

(22)

sensitivity to the distribution of ‘fake’ reflexives in English, correctly accepting (a) examples and rejecting (b) examples below:

(1.1) a. Calling for her dog, Sue shouted herself hoarse.

* b. Calling for her dog, Sue shout hoarse.

(1.2) a. Before Brian could smooth it off, the plaster dried hard.

*b. Before Brian could smooth it off, the plaster dried itself hard.

Hawkins et al. (2006) argue that the distribution of fake reflexives is governed by syntactic constraints on the realization of argument structure, which is related to a Case requirement. Specifically, adjectives used as resultative complements (hoarse and hard) require “an appropriately positioned specifier” (Hale and Keyser 2002:5). Unaccusative verbs like dry contrast with unergative verbs like shout in the availability of such a specifier position in their respective argument structure. Such a specifier position is available in the argument structure underlying unaccusative verbs, hence no further argument need be projected in the derivation of the sentence; it is absent in the argument structure underlying unergative verbs, hence the resultative complement must project one, realized in English as a ‘self’ form. These constraints are invariant across languages that allow resultative constructions, hence are assumed to be a universal principle. The distribution of fake reflexives in English meets the three conditions designated by White (2003:23) for ‘poverty-of-the-stimulus’ instance in L2: (a) the property in question should not be available in the L1; (b) it must be underdetermined by input; (c) it should not be known through instruction or other metalinguistic means.

Thus the L2 learners’ knowledge of the distribution of fake reflexives in English demonstrates that adult L2 learners indeed have access to UG principles. This entails that UG principles are not subject to a critical period.

As to the correlation between age effects and parameters, no consensus has been reached yet. L2 researchers are still debating whether or not UG-based parameters are affected by a critical period. Two opposing positions emerge in the current SLA literature. One position, represented by proponents of Failed Functional Features Hypothesis (FFFH, e.g. Tsimpli & Roussou 1991; Hawkins and Chan 1997; Hawkins 2001b; 2004; Franceschina 2001; Hawkins and Liszka 2003; Tsimpli 2003), claims that (at least part of) parameterized properties are subject to a critical period, leading to representation deficits in L2 interlanguage grammars (hence it is also referred to as Representational Deficit Hypothesis (RDH)). The other position, represented by L2 researchers espousing Full Transfer Full Access Hypothesis (FTFA, e.g. Schwartz and Sprouse 1996; White 2003), maintains that new parameters can be reset to settings of the target language even after a critical period, and that L2 speakers have the same underlying grammatical representations as native speakers. What is vulnerable to maturational constraints is the more peripheral components of the language faculty, such as the morphological module (termed the Missing Surface Inflection Hypothesis (MSIH), see for example, Lardiere 1998a, 1998b, 2000, 2006a; Prévost and White 2000) and the phonological module (termed the Prosodic Transfer Hypothesis (PTH), see for example, Goad et al. 2003; Goad and White 2004; Goad and White 2006). These different approaches to explaining NS-NNS divergence will be evaluated in the current study in relation to the L2 data from the experiments.

(23)

My study also explores another fundamental question concerning L2 ultimate attainment. This question regards whether endstate L2 speakers can attain nativelike processing capacity. This issue has not received much attention in the previous research on L2 ultimate attainment, probably because processing performance and grammatical competence have been viewed as separate subjects and belonging to different realms of research. This is a misconception. As pointed out by Hyltenstam and Abrahamsson (2003:576), it is a ‘problem’ that L2 researchers doing ultimate attainment studies tend to equate ‘language’ with ‘grammatical competence’. They further remark that “the domain for maturational constraints in general is the human capacity for language both at the level of knowledge and at the level of processing” (2003: 577). I fully agree with them on this point, and believe that our explanation for L1-L2 divergence will never be adequate if we do not link second language acquisition studies and psycholinguistic experimentation. In the present study, I follow the view of Hyltenstam and Abrahamsson and adopt a comprehensively defined notion of linguistic competence, including both grammatical competence and processing capacity. Investigating maturational constraints both at the level of knowledge and at the level of processing constitutes the primary goal of the current study.

The target language in this study is Norwegian, a language, together with Swedish, Danish, Icelandic, and Faroese, known as a Scandinavian language. To locate Norwegian in an upper branch of the language family tree, we find that it belongs to the Germanic language family, which makes it a close kin to English. But in fact Norwegian and English differ in many important aspects, including the syntactic patterning and the main inflectional categories, especially in the nominal domain. As we will discuss at length in Chapter 2, the Norwegian language is characterized by a rather complex inflectional system in its nominal domain, and there is agreement between the various constituents within nominal determiner phrases (DP). Norwegian DP internal agreement will be the focus in the current endeavor to investigate L2 grammatical knowledge and processing capacity. The topic was chosen because it has been shown that L2 speakers often have persistent difficulty with agreement morphologies (e.g.

Long 1997; Lardiere 1998a, 1998b, 2006b; White 2002; Hawkins & Franceschina 2003). The potential area where adult L2 acquisition failure is likely to occur will make a good testing point for the various theoretical constructs seeking to explain L2-L1 divergence.

Acknowledging that “the most fruitful way to research maturational constraints is to focus explicitly on ultimate L2 learning” (Hyltenstam and Abrahamsson 2003), the current study investigates L2 knowledge and L2 processing capacity with respect to Norwegian DP internal agreement by examining the eventual outcomes of three groups of L2 learners. Since these informants used L2 Norwegian, along with their L1s, on a daily basis for a sufficient period of time (see Section 2 of Chapter 3 for details), they were virtually L2 speakers (or more generally L2 users), rather than L2 learners. The L2 speakers were selected from typologically different L1 backgrounds: Italian/Spanish, English, and Chinese. Syntactically, as we will discuss in some detail later (in Chapter 3), the three source languages differ parametrically from the target language with respect to one, two, or three of the uninterpretable features relevant for DP internal agreement. Thus, investigating the ultimate attainment of the three groups of L2 speakers will allow us to explore “the role of the L1 in delimiting what is ultimately attainable in the L2” (Lardiere 2006c:240).

(24)

It ought to be pointed out that the Norwegian language has rarely been the subject for generative second language acquisition research, contrasting its significant contributions to generative syntactic theories.² This seems to be a general trend for all Scandinavian languages, as can be seen from the following remark by Josefsson and Platzack (2004:1):

Although the Scandinavian languages have played an important role as research object in the development of generative grammar during the last twenty years, they have not been a prominent subject for studies of language acquisition within this research paradigm (Josefsson and Platzack 2004:1).

This situation is especially true for Norwegian. To my knowledge, the current study will virtually be the first doctoral dissertation on the acquisition of Norwegian as a second language embedded in the Principles and Parameters framework. It thus represents the initial efforts towards narrowing the gap of the contributions Norwegian has made to syntactic theories and to empirical SLA research. Through this study, I wish to demonstrate that the Norwegian language offers some special opportunities for the investigation of second language acquisition, especially regarding L2 ultimate attainment. Indeed L2 research requires the pairing of diverse source and target languages in order to gain a more comprehensive understanding about the nature of SLA, and about the nature of “the complex system of knowledge that has actually been attained” (Chomsky 1986).

1.2 Research questions

The present study addresses three main research questions with a view to investigating L2 grammatical competence and processing capacity. L2 knowledge with respect to Norwegian DP internal agreement is firstly investigated through an on-line DP production task, as on-line tasks are generally believed to be reliable measures of implicit linguistic knowledge (see for example, Ellis 2005). In addition, L2 learners’

knowledge about the Norwegian gender system is investigated separately, using a gender assignment task, as a gender feature is inherent in Norwegian nouns (contrasting, for example, number, which is a derived feature). These two experiments (which will also be referred to as L2 production tasks) together address the research question RQ1:

• RQ1: Can L2 speakers attain nativelike success in grammatical competence with respect to Norwegian DP internal agreement? If not, which of the current proposals (the FFFH/RDH vs. the MSIH or the PTH) is most compatible with the L2 data?

L2 processing will be tested using an auditory naming technique. As no previous study has been conducted on native speakers’ processing of Norwegian DP internal agreement, L1 processing performance will be investigated along with L2 processing.

The auditory naming task can presumably reveal to us whether L1/L2 speakers can perceive the (concordant and discordant) gender/number/definiteness agreement cues

2 There is ample research related to the syntactic structure of nominal determiner phrases alone, from the perspective of Norwegian or Scandinavian languages in general, see among others Taraldsen 1990;

(25)

available on the determiners (hence it will also be referred as the L2 perception task).

This experiment examines research question RQ2:

• RQ2: Do endstate L2 speakers process Norwegian DP internal agreement the same way as the natives do?

Results from the two production experiments can give valuable insights on designing research questions for the subsequent study on processing performance. If some L2 speakers are found to perform in a nativelike way in the on-line production task, it is of interest to compare the processing behaviors of nativelike performers (those who presumably have attained nativelike grammatical competence) and non-nativelike performers (those who presumably have not attained nativelike grammatical competence) and see in what way L2 processing behaviors are related to L2 grammatical competence. Exploring the interrelation between L2 production and L2 perception gives rise to the third research question RQ3:

• RQ3: Is L2 speakers’ nativelike (or non-nativelike) processing directly related to their L2 grammatical competence? In other words, does perception parallel production?

I am not aware of any previous research on ultimate attainment (e.g. Long 1997;

Lardiere 1998a, 1998b, 2006b; Franceschina 2001; White 2002) addressing RQ3. This current study is probably the first to investigate the interrelation between L2 speakers’

grammatical competence and syntactic processing capacity. At present, all the three research questions are of a general nature; they will be developed and refined in the subsequent chapters where the experiments are reported.

1.3 Thesis outline

This chapter briefly presents the research contexts and sets the background for the three main research questions in the current study. The remaining chapters are structured as follows. Chapter 2 gives a descriptive overview of the morphosyntax in the nominal domains of the target language and the source languages. As this study deals with the acquisition and processing of DP internal agreement, this chapter gives a comparative description of how DP internal agreement, if any, is realized in the languages under discussion. In Chapter 3, I present the analyses of DP structures of the target language and the source languages, based on the DP model proposed by Julien (2003, 2005).

Parametric differences between the target and the source languages with respect to uninterpretable features, which are presumably susceptible to maturational constraints, are discussed at some length. As the current study and Julien’s model are both conducted within the Minimalist framework, this chapter also includes in its initial section a brief introduction to the Minimalist Program, in particular its basic concepts and major claims, which are important for understanding the theoretical issues under investigation in this study. Chapter 4 is a review of SLA theories, including both theories on grammatical competence and theories on L2 processing. For the former part (i.e. theories on grammatical competence), the focus is on what I take to be the two competing SLA models within the generative framework, namely, the Full Transfer Full

(26)

Access hypothesis (FTFA) and the Failed Functional Features Hypothesis (FFFH). I will show that the two models represent the two opposing positions regarding age effects on parameter resetting, and as such are especially relevant to studies of L2 ultimate attainment. Their different views on maturational constraints and different approaches accounting for NS-NNS divergence will be presented and assessed. For the latter part (i.e. theories on L2 processing), a L2 processing theory known as the Shallow Structure Hypothesis (SSH, Clahsen and Felser 2006a, 2006b) will be outlined. In Chapter 5, I start with a detailed description of the L2 participants, as well as their scores achieved on the Norwegian proficiency test, then go on to report on two experiments which aim at ‘tapping’ L2 grammatical competence. The first experiment is an off-line gender-assignment task, where participants were asked to assign the correct gender to the Norwegian nouns; the second experiment is an on-line DP production task, through which Norwegian DPs containing adjectives were elicited. The results will be subsequently presented, and the extent to which the FTFA and the FFFH models can satisfactorily account for the empirical data will be discussed. The non-target-like data will be used to evaluate the various approaches (namely, the FFFH/RDH, the MSIH, and the PTH) to explaining NS-NNS divergence. Chapter 6 reports on an experiment exploring L1 and L2 processing behaviors. First an overview of the relevant psycholinguistic studies is provided, followed by a discussion of agreement processing mechanisms and a proposal for an agreement processing model. L1 and L2 results are subsequently presented and compared. A comparison is also made between the three L2 groups, and between the nativelike performers and non-nativelike performers in the on- line production task. This chapter ends with a discussion of L1 and L2 results in relation to current processing theories and implications of the finding for the SSH. Finally Chapter 7 provides answers to the research questions, summarizes the contributions of the thesis, and points out directions for future work.

(27)

2 Morphosyntax of DPs in the target and the source languages

2.1 Introduction

In this Chapter, I will deal with the structural aspect of the determiner phrases of the target language, Norwegian, and of the informants’ first languages, namely, English, Italian/Spanish, and Chinese.³ Languages often differ as to what grammatical categories they encode, and as to the way a certain category is encoded. Among the three source languages, only Italian/Spanish has inflectional categories of gender, number, and definiteness in its nominal domains, as in the target language. Neither English nor Chinese has a gender system. While English encodes both number and definiteness, Chinese, as an isolating language, generally lacks gender, number, and definiteness distinctions, though suffix –men is arguably considered as a plural marker in some literature. The aim of this Chapter is to give a descriptive overview of the morphosyntax in the nominal domain of the languages under discussion. After a brief comparison of the nominal systems of the four languages in Section 2.2, I look at the similarities and differences between the Norwegian and Italian/Spanish gender and number systems and present a detailed description of how DP internal agreement is realized in those two languages in Section 2.3. A summary follows in Section 2.4.

2.2 A comparison of DPs of the target and source languages

In this section, I will make a general comparison of DPs in the target and source languages. Specifically, I will discuss the following three points: 1) whether a certain grammatical category has overt morphological reflexes in the DPs of L2 Norwegian and the source languages; 2) how a specific grammatical category is encoded, and 3) whether the languages exhibit DP internal agreement in terms of gender, number, or definiteness. This is followed by a discussion of whether there is a number agreement in Chinese and whether Chinese has number morphology at all.

3 In the present study, the term Chinese refers exclusively to Mandarin Chinese. Only the properties of Mandarin Chinese are discussed in this dissertation, and the Chinese informants in this study all acquired Mandarin Chinese as their native language.

(28)

As far as the nominal domain is concerned, the target language, Norwegian, has a relatively complex inflectional system, which encodes gender, number, and definiteness. And gender, number, and definiteness agreement (concord) is mandatory between nouns and determiners, nouns and adjective modifiers.⁴ This is illustrated in (2.1) below.⁵

(2.1) a. et fin-t land

D-NEUT.SG.INDEF nice- NEUT SG.INDEF country-NEUT.SG.INDEF

‘a fine country

b. de fin-e land-ene

D-PL.DEF nice- DEF country-NEUT.PL.DEF

‘the nice countries’

c. fin-e bil-er

nice-PL car-COM. PL.INDEF ‘nice cars’

Note specially that in (2.1b) there is a co-occurrence of a free, prenominal determiner (de) and a suffixed determiner (-ene) on the noun. This construction is traditionally called a “double definiteness” construction. “Double definiteness” is a unique feature of the Norwegian language, and will be elaborated in the subsequent section.

Among the source languages, Italian/Spanish is the only language that has gender, number and definiteness inflections, and gender and number agreement must be marked on determiners, adjectives, and all coreferential pronouns. But unlike Norwegian, there is no definiteness agreement in Italian/Spanish. Some Italian examples are provided below to illustrate this point.

(2.2) a. un libro carino D-MASC.SG. INDEF book-MACS.SG nice-MASC.SG.

‘a nice book’

b. il libro carino D-MASC.SG.DEF book- MASC.SG nice- MASC.SG

‘the nice book’

c. libri carini book- MASC.PL nice- MASC.PL

‘nice books’

English has a relatively simple nominal system, with inflections for number (-s as plural marker on the nouns; singular number is unmarked, and therefore not visible on

4 In addition, gender and number agreement is found between subjects and predicative adjectives. But this agreement relation is not of concern in the present study.

5 The following symbols are used for the annotation throughout the paper. D stands for prenominal determiners; NEUT for neuter gender, COM for common gender, MASC for masculine gender, F for

(29)

the nouns). Moreover, English has number agreement between nouns and demonstratives, but not between definite articles and nouns, nor between adjectives and nouns, as shown in the following set of examples below:

(2.3) a. the nice cup the nice cups

b. this/that nice cup these/those nice cups

In comparison to most European languages, gender plays a rather small part in the grammar of English. With a few exceptions (e.g. actor vs. actress; waiter vs. waitress), English nouns generally do not encode a gender distinction. Accordingly, English does not exhibit gender agreement. Nonetheless the third person singular pronouns require the distinction between masculine he, feminine she, and neuter it, and there is gender agreement between nouns and their pronouns.

Among the languages in question, Mandarin Chinese is the only non-Indo-European language. It is known as an isolating language, a language without much inflectional morphology. In the nominal domain, Mandarin Chinese is first and foremost characterized by a total absence of grammatical gender: there is even no gender distinction for third person singular pronouns in speech, where ta serves as translation equivalent of all three English pronouns, though in the orthographic system they are distinguished according to masculine 他, feminine她, and neuter gender它.

The second striking feature of Chinese DPs is lack of number and definiteness markers; there is no morphological clue showing whether a noun phrase is definite or indefinite, singular or plural. Instead, a ‘bare’ noun can appear in any argument position, the number and definiteness of its referent being unspecified, as in (2.4):

(2.4) Ta qu na pingguo.

he/she go get apple

‘He/She went to get an apple/the apple/apples/the apples.’

A bare noun such as pingguo ‘apple’ in the above sentence can be interpreted as eithersingular or plural, definite or indefinite, the context providing the necessary information. Indeed, Chierchia (1998) proposes that Chinese NPs are all mass nouns in nature; mass nouns are ‘argumental’, and can be merged directly in argument positions in syntactic expressions without the need for a determiner. This is in contrast with Spanish/Italian (presumably Norwegian and English as well), whose NPs are

‘predicative’, hence can never be merged directly in argument positions, but require a determiner to license them as arguments.

Chinese is also known as having a rich classifier system. Classifiers “classify”

referents of the nouns into a certain category based on some salient features of the objects that the nouns denote. For example, the classifier ben is used for books and magazines, whereas zhang is used for rectangular, flat objects such as tables, beds, maps. If a noun does not have its own specific classifier, then it takes ge, which is used as a general classifier. Syntactically, the classifiers have been argued to have “an individualizing or singularizing function” (Chen & Sybesma 1999), through which nouns become “syntactically visible and countable” (Doetijes, 1996). Hence a classifier is obligatory when the number of the individuals that a noun denotes is indicated, as we can see from the grammaticality contrast below:

(30)

(2.5) a. san ge xuesheng b. * san xuesheng three CL student three student ‘three students’ ‘three students’

With respect to the number feature in Mandarin Chinese, there is a great controversy.

Traditionally, the suffixes -men and –xie have been considered plural markers.

However, nowadays there has been no general consensus on the status of the two suffixes, due to their distinct behavior from the plural marker in European languages and their co-occurrence constraints.

Unlike the English plural morphology –s, -men cannot be attached to all countable nouns; its distribution is restricted to animate nouns and personal pronouns. In addition, -men cannot occur with numerals, whether or not a classifier is present. This can be observed from the ungrammaticality of the following expressions:

(2.6) a. * san ge xuesheng-men b. * san xuesheng-men three CL student three student Intended meaning: ‘three students’

As for –xie, a number of linguists (see for example, Lü 1980) consider it the plural marker for demonstratives zhe ‘this’ and na ‘that’, forming zhe-xie ‘these’ and na-xie

‘those’; others (e.g. Li & Thompson 1989; Chen & Sybesma 1999) treat it as an indefinite plural classifier. The issue which is of concern here is whether there is number agreement between the demonstratives and nouns in Chinese, as seemingly an agreement relation exists in the following Chinese noun phrases (for expository purposes, I temporally use PL for both –men and –xie in the notation):

(2.7) a. zhe –xie xuesheng(-men) b. na-xie xuesheng (-men) this-PL student-PL that-PL student-PL

‘these students’ ‘those students’

If it is indeed the case that –xie marks the plurality of the demonstratives, and –men the nouns, Chinese should have the same number agreement relation as English, the only difference being that the number suffix –men in Chinese is optional, while –s in English is obligatory. Indeed this is the claim made by Kuroda (1988): number agreement is obligatory in English, optional in Chinese.

But any claim for –men and -xie as plurality markers for nouns and demonstratives, respectively, would have to explain why neither -men nor –xie can co-occur with a numeral (>1). The co-occurrence constraint of –xie is illustrated by the contrast below (Also see examples in (2.6) for the co-occurrence constraint of –men):

(2.8) a. zhe san ge xuesheng b. * zhe-xie san ge xuesheng this three CL student this-PL three CL student-PL

‘these three students’ ‘these three students’

The examples in (2.8) suggest that the singular demonstrative can co-occur with numerals (>1), whereas the plural demonstrative cannot—a fact that is in reverse to those languages that display number agreement, such as English, Norwegian. To

(31)

account for this puzzle, I follow Yang (2004) in arguing that –men is a suffixal classifier instead of a plural marker, and that xie is a plural element under Num⁰. Under this account, the demonstrative phrases zhe san ge xuesheng ‘these three students’ and zhexie xuesheng (-men) ‘these students’ have a syntactic representation in (2.9a) and (2.9b), respectively:⁶

(2.9) a. zhe san ge xuesheng ‘these three students’

DP

D NumP zhe

‘this’ Num CLP san

CL NP ge xuesheng [-group] ‘student’

b. zhexie xuesheng ‘these students’

DP D NumP zhe

‘this’ Num CLP xie

[number: pl] CL NP -men xuesheng [+group] ‘student’

Yang assumes that the suffixal classifier –men has a [+group] feature, which gives rise to a group interpretation of an NP. She further argues that the lexical item xie is the overt realization of the plural feature under the Num head. Numerals themselves, however, do not have a plural feature. Since both xie and numerals are elements on the Num head, they compete for the same syntactic position, hence the two cannot co-occur.

This accounts for the co-occurrence constraint between xie and numerals (>1). After NP merges with CLwith a [+group] feature as illustrated in (2.9b), N head moves to CL to pick up the [+group] feature of –men. The next step when Num is merged, it must contain a [+pl] feature, otherwise there will be a feature clash between the [+group]

feature on CL+N and the [-pl] feature on Num. If Num has a plural feature, under Yang’s assumption, the plural feature has to be realized as xie. On the other hand, if NP merges with CL with a [-group] feature, as illustrated in (2.9a), numerals which are without a plural feature are compatible with the [-group] feature on CL. Crucially, D in Chinese does not have a number feature, so in the above two cases, zhe ‘this’ is spelt out irrespective of the number features on the nouns.

In short, Yang proposes that 1) –men is a suffixed classifier, instead of a plural marker, 2) xie is a plural element under Num, instead of a plural suffix of

6 The syntactic structure of Chinese DP presented here is based on Yang (2004), though it is subject to modification in the subsequent chapter.

(32)

demonstratives, and 3) D in Chinese does not involve a number feature. Her proposal, in my view, nicely captures the co-occurrence constraints of xie and –men with numerals (>1). More relevant to my discussion is the conclusion that Chinese does not show D-N agreement in terms of number.

To summarize so far, we have discussed whether the languages in question have overt morphology in terms of gender, number, and definiteness, and whether they exhibited DP-internal agreement with respect to these features. As can be seen from Table 2.1, the target language, Norwegian, has overt morphological reflexes of number, gender and definiteness. It also has overt agreement between determiners, adjectives, and nouns in terms of number, gender and definiteness. Italian/Spanish share all the nominal properties with Norwegian except for definiteness agreement, while Chinese is distinct from Norwegian in that it has no overt morphological expression for all these nominal properties. English, like Chinese, has no gender or gender agreement in its nominal domain, but it encodes number and definiteness, and it exhibits overt number agreement between demonstratives and nouns.

Table 2.1. Inflectional categories and agreement in DPs of the L2 and the L1s Norwegian Italian/Spanish English Chinese

Number + + + -

Number agreement + + + -

Gender + + - -

Gender agreement + + - -

Definiteness (DEF) + + + -

DEF agreement + - - -

Key: + Present in language; - absent in language

2.3 Gender and number systems of Norwegian and Italian/Spanish

As has been mentioned before, Italian/Spanish is the only source language, like the target language Norwegian, which has gender and number agreement. But as members of different language families (namely, Germanic and Romance), Norwegian and Italian/Spanish have structurally distinct gender and number systems. Since the structure of gender and number systems in the first language presumably has some effect on L2 acquisition and perception, this section will be devoted to a detailed discussion of the similarities and differences between the gender and the number systems in the two languages. The gender and the number systems in English and Chinese are much simpler, and have been touched upon in the previous section, so they are not subjected to further discussion in this section.

2.3.1 Gender and number in Norwegian

Norwegian is a language with a more formal gender system than a semantic gender system (see Corbett 1991, for a distinction between formal and semantic gender systems). Generally, the gender of the nouns bears no relation with the sex of nouns’

referents. For example, lærer ‘teacher’ is assigned a masculine gender, irrespective of

(33)

the fact that it refers to a male or a female teacher.⁷ In many Norwegian dialects, there are three genders: neuter, masculine and feminine. The three genders are not evenly distributed, though. According to the Oslo Corpus of Tagged Norwegian Text (website address: http://www.hf.uio.no/tekstlab/frekvensordlister/index.html), the frequency of occurrences for the three indefinite articles in magazines and newspapers are 126870 for en (masculine indefinite), 57635 for et (neuter indefinite), and 957 for ei (feminine indefinite). This means masculine nouns make up 68% of all the indefinite singular nouns in the corpus, neuter nouns 31%, and feminine nouns only 1%. The small proportion of feminine nouns can be (partly, at least) attributed to the fact that in Bokmål Norwegian, masculine and feminine genders can be collapsed into a common gender (en), which is in the same form as the masculine gender. For example, jente

‘girl’ is a feminine noun (e.g. ei jente) in the three way gender system. In the two way system, it has a common gender (i.e. en jente) instead. Actually in the three-way gender system, the distinction between masculine and feminine gender can only be visible on singular indefinite articles, and optionally on the suffixed definite article on nouns; the agreement inflections on the prenominal definite articles and adjectives are exactly the same for both genders, as shown in the two sets of examples in (2.10), illustrated by a feminine noun jente ‘girl’ and a masculine noun bil ‘car’. In order to simplify assumptions, this paper adopts the binary gender system of Norwegian. That is to say, Norwegian nouns distinguish a neuter gender from a common (non-neuter) gender.

Common gender, which is over twice as frequent as neuter gender, is considered as the default or the unmarked form, whereas neuter gender is the marked form.

(2.10) a. ei fin jente den fine jenta/en de fine jentene ‘a fine girl’ ‘the fine girl’ ‘the fine girls’

b. en fin bil den fine bilen de fine bilene ‘a fine car’ ‘the fine car’ ‘the fine cars’

Gender assignment often seems to be largely arbitrary in Norwegian. In addition, phonological forms of nouns give no cue to their gender.⁸ For example, ball ‘ball’ is assigned a common gender, while tall ‘number’ a neuter one; brann ‘fire’ is assigned a common gender, while våpen ‘weapon’ a neuter one; kirke ‘church’ is assigned a common gender, while ønske ‘wish’ a neuter one.

Cues to gender classification of nouns are exhibited in determiners, attributive adjectives, and various other elements that co-occur with the head noun. In singular indefinite neuter context, gender (and number) agreement has an overt phonological realization both on the determiner and on the attributive adjective, while in the corresponding common gender, it has a phonological realization only on the determiner, but not on the attributive adjective (the null morpheme is marked with a zero symbol

‘ø’). This is illustrated in (2.11) below.

7 There is an antique form lærerinne for female teachers, but this form is rarely used in modern Norwegian.

8 Certain derivational suffixes, however, may provide some morphological cues to gender. Suffixes, such as –sjon, -else, -ning, -het, -dom, -nad, are associated with common gender, whereas –ment, -eri, - em, -tek are associated with neuter gender.

(34)

(2.11) a. et fin-t land

D-NEUT.SG.INDEF nice- NEUT country-NEUT.SG.INDEF

‘a fine country’

b. en fin bil

D-COM.SG.INDEF nice- COM(Ø) car-COM.SG.INDEF

‘a nice car’

Also note that the realization of –t, which marks neuter indefinite on adjectives, is rather restricted. Generally adjectives ending in 1) –t or -d, and 2) in a derivational suffix, remain invariant. Actually such adjectives are very pervasive, and are actively involved in everyday life. Adjectives from the first category include lett ‘easy’, fast

‘steady’, glad ‘happy’, opptatt ‘busy’; from the second category, vanskelig ‘difficult’, tidlig ‘early’, viktig ‘important’, alvorlig ‘serious’, norsk ‘Norwegian’, amerikansk

‘American’, to name just a few. This means in many cases this gender agreement is not visible on the adjectives.

More complication is involved in Norwegian definite determiner phrases. Unlike indefinite DPs, where the indefinite determiner is a prenominal free morpheme, the definite determiner is realized as a suffixed bound morpheme, which encodes gender and number information as well. What is more intriguing, when an attributive adjective (or a numeral) is inserted in the definite DPs, it requires the co-occurrence of both a suffixed determiner and a free prenominal determiner. Hence, when modified, the Norwegian definite DPs exhibit a definiteness agreement between nouns and determiners, apart from gender and number agreement. Due to what appears to be double marking of definiteness, this construction is commonly referred to as a ‘double definiteness’ construction.⁹ In contrast to the ‘strong’ inflection in indefinite DPs, where adjectives show number and gender distinctions, the adjectives in the double definiteness construction take an invariant suffix –e. This invariant form is often referred to as weak inflection, and I use w to stand for it in annotation. The following are examples illustrating the agreement paradigm of Norwegian definite DPs:

(2.12) a. land-et country- NEUT.SG.DEF

‘the country’

b. bil-en car-COM.SG. DEF ‘the car’

c. det fin-e land-et

D-NEUT.SG.DEF nice-w country-NEUT.SG.DEF ‘the nice country’

d. den fin-e bil-en

D-COM.SG.DEF nice-w car-COM.SG.DEF

9 Apart from Norwegian, the ‘double definiteness’ construction can be found in Swedish and Faroese,

(35)

‘the nice car’

In addition, Norwegian demonstrative phrases also require double definiteness, irrespective of presence or absence of a modifier. Demonstratives are inflected for gender and number. Besides, they also distinguish on the proximity. Examples of demonstrative double definiteness are given below. Note that the distal forms of demonstratives resemble the free prenominal determiners.

(2.13) Distal

a. det (fin-e) land-et that-NEUT.SG. (nice-w) country- NEU.SG.DEF

‘that (nice) country’

b. den (fin-e) bil-en that-COM.SG. (nice-w) car-COM.SG.DEF

‘that (nice) car’

(2.14) Proximal

a. dette (fin-e) land-et this-NEUT.SG. (nice-w) country- NEUT.SG.DEF

‘this (nice) country’

b. denne (fin-e) bil-en this-COM.SG. (nice-w) car-COM.SG.DEF

‘this (nice) car’

It should be pointed out that this ‘double marking’ of definiteness in modified Norwegian DPs is not obligatorily required in all cases. ‘Single definiteness’, where either the suffixed determiner or the prenominal determiner is left out, is either optional or obligatory, depending on the contexts and various other factors. The prenominal determiner is often left out in nominals which are deictic (cf. example 2.15a), vocative (cf. example 2.15b), depictive (cf. example 2.15c), or are associated with a proper name (cf. example 2.15d) (Julien 2005: 30-34; Delsing 1993:116-120).

(2.15) a. Ta (den) ny-e bil-en. (adapted from Delsing 1993:118) take (the) new-w car-COM.SG.DEF

‘Take the new car.’

b. Vet du ikke det, (*den) stor-e jent-en! (Julien 2005:32) know you not that (D- COM.SG.DEF) big-w girl-COM.SG. DEF

‘Don’t you know that, you big girl!’

c. Det var (*den) svart-e natt-en da hun kom. (ibid) it was (D-COM.SG.DEF) black-w night-COM.SG.DEF when she came

‘It was dark night when she came.’

d. Vi besøkte Hvit-e Hus-et. (ibid)

(36)

we visited white-w house-NEUT.SG.DEF

‘We visited the White House.’

On the other hand, the suffixed determiner can be left out if the nominals get a nonspecific reading (Julien 2005:35-39). Julien points out the specific/nonspecific contrast using the following examples:

(2.16) a. Jeg lik-er det hvit-e gull-et bedre.

I like-PRES D-NEUT.SG.DEF white-w gold-NEUT.SG. DEF better ‘I like the white gold better.’

b. Olje kalle-s iblant ‘det sort-e gull’.

oil call-PASS sometimes D-NEUT.SG DEF black-w gold ‘Oil is sometimes called ‘the black gold’.

The DP in example (2.16a), det hvite gullet, has a specific reading; whereas the DP in (2.16b), the one without the suffixed determiner on the noun, gets a non-specific reading. What is NOT mentioned in the literature is that the suffixed determiner can also be left out in a formal speech or written discourse, without any semantic effect (Åfarli 2005, personal communication). In a sense, the drop of the suffixed determiner in the ‘should-be’ double definiteness construction simply signals the formal genre of the discourse. This particular usage to a certain extent reflects the residue of Danish influence on the Norwegian language. Danish resembles Norwegian within DPs in all other aspects except that it does not have the ‘double definiteness’ construction, as one can observe from the following set of Danish DP examples:

(2.17) a. et land

D-NEUT.SG. country-NEUT.SG.

‘a country’

b. land-et country- NEU.SG.DEF

‘the country’

c. et fin-t land

D-NEUT.SG. nice- NEUT country-NEUT.SG.

‘a fine country’

d. det fin-e land

D-NEUT.SG. nice-w country-NEUT.SG. ‘the nice country’

We see that in Danish, there is a pronominal determiner and no suffixed article in definite DPs containing an attributive adjective. The Bokmål Norwegian (which used to be called Dano-Norwegian) was actually developed on the basis of Danish, while integrating into it “specifically Norwegian elements from the colloquial speech of the

(37)

educated classes in urban areas (Askedal 1994).” It is thus natural that Norwegian retains some traces of Danish syntactic norms in its formal style of speech and writing.

With respect to number, Norwegian has singular (SG) and plural (PL) distinctions.

Singular number is the unmarked form (or zero form), while plural is the marked form.

Plural is generally expressed with a suffix, except for monosyllabic neuter nouns in indefinite context, which remain invariable in the plural (cf. example 2.18a). When the noun is plural, gender information is not visible on the plural suffixes, which are –er and –ene for indefinite and definite respectively. This is illustrated in the following set of examples:

(2.18) a. (mange) land [neuter, monosyllabic]

(many) country-PL(Ø).INDEF

‘many countries’

b. land-ene country-PL. DEF

‘the countries’

c. (mange) bil-er [common gender, monosyllabic]

(many) car- PL.INDEF ‘many cars’

d. bil-ene car- PL. DEF

‘the cars’

In addition, adjectives and prenominal determiners all have plural forms. The plural form of the prenominal determiner is de, irrespective of the gender of the following noun. Plural indefinite DPs do not require a prenominal determiner, or put another way, they require a null prenominal determiner (which is symbolized with a zero marker ø).

The plural form of adjectives takes an invariant suffix –e, both in definite and indefinite contexts. Modified plural Norwegian DPs are illustrated in (2.19):

(2.19) a. ø fin-e land nice-PL country-NEUT.PL (Ø).INDEF

‘nice countries’

b. de fin-e land-ene

D-PL.DEF nice-w country- NEUT.PL.DEF

‘the nice countries’

c. fin-e bil-er

nice-PL car-COM. PL.INDEF ‘nice cars’

d. de fin-e bil-ene D- PL. DEF nice-w car-COM.PL.DEF

‘the nice cars’

(38)

To sum up, I have presented a rather comprehensive overview of Norwegian DPs and realizations of DP-internal agreement. Table 2.2 lists the regular inflectional paradigm of modified Norwegian DPs with all possible feature combinations.¹⁰

Table 2.2. Inflection paradigm of modified Norwegian DPs

FEATURE BUNDLES D ADJ N

[SG][NEUT][INDEF] et -t -ø [SG][COM][INDEF] en -ø -ø

[SG][NEUT][DEF] det -e -et

[SG][COM][DEF] den -e -en

[PL][INDEF] ø -e -er

[PL][DEF] de -e -ene

As we see, (the prenominal) D has altogether six (free) morphological forms; ADJ has three realizations: -t marks neuter singular indefinite; zero (-ø) marks common singular indefinite; -e marks the rest. N has five inflectional forms, zero (-ø), –et, –en, - er, and -ene. The irregularities of inflectional system and context-dependent omission of determiners in the double definiteness construction add further complexity to this picture. But the present study focuses on the regular inflection paradigm. The inflectional paradigm in Table 2.2 above thus provides an important reference point in assessing L2 learners’ success with respect to DP internal agreement.

2.3.2 Gender and number in Italian/Spanish

As is the case for all Romance languages, Italian and Spanish have gender and number distinctions, and both languages have DP internal agreement in terms of gender and number. Italian and Spanish gender and number systems have a lot in common, so for the purpose of this study I consider them as one system. At the points where they differ from each other, I will explicitly point out.

Noun gender in Italian and Spanish falls into two categories: masculine and feminine. This is the first point where Norwegian and Italian/Spanish deviate. The two way gender distinction in Italian/Spanish, corresponding to masculine and feminine, do not map into the two way gender system of Norwegian, corresponding to common gender and neuter. For example, Italian nouns mela ‘apple’ and palazzo ‘building’ are feminine and masculine respectively, whereas their Norwegian counterparts, eple and bygg, are both of neuter gender. The second point where Italian/Spanish and Norwegian gender systems differ is that in Italian and Spanish, gender is to a large extent transparently marked on word endings. The general rule is that nouns ending in –o are masculine; nouns ending in –a are feminine. To use Bates et al’s (1996) metaphor, the noun ‘wears its gender on its sleeve’. In contrast, the phonological forms of Norwegian nouns generally provide no cue to their gender. So learners of Norwegian cannot rely on word endings; rather they have to learn the gender on a word-by-word basis. Of course, there are nouns whose genders are not predictable on the basis of word endings in Italian/Spanish as well. For example, Italian nouns ending in –e can either be masculine

10 The inflection system presented here is based on the Bokmål variety of Norwegian. There are dialectal variations in this inflectional paradigm.

Second Language Acquisition and Processing of Norwegian DP Internal Agreement